System and Method of Adjusting Smoothing

The invention pertains to ambient condition detectors. More particularly, the invention pertains to such detectors which incorporate variable smoothing of signals indicative of an ambient condition.

Distributed fire alarm systems which incorporate a plurality of ambient condition detectors, such as smoke, heat or gas detectors, are often installed in business or commercial buildings. Such systems often have a common control unit which can be in either unidirectional or bidirectional communication with multiple, spatially separated, ambient condition detectors.

One of the problems associated with transmission of information to or from such detectors is the presence of uncorrelated noise. Noise is uncorrelated wherein it is not related to a selected parameter or parameters which is/are being monitored.

In the event that the parameter being monitored is a level of ambient smoke, an ambient temperature, or a level of an ambient gas, the signals of interest are those which have a high correlation to the particular ambient condition being detected. Other signals, due to electrical or thermal noise which are not correlated to the ambient condition being detected, and which may in fact be random, are undesirable. Various techniques have been used in the past to minimize the effects of such uncorrelated noise signals.

One known type processing or filtering involves sampling the signals from at least one of the ambient condition detectors and calculating a running average based on a predetermined number of prior sample values, such as 6 or 8 or 10, along with the latest sample value. As each new sample value is received, the running average is updated. This technique provides a vehicle for minimizing or suppressing the effects of uncorrelated noise. This process can also be carried out continuously using analog circuits.

Filters can be implemented using analog or digital hardware. Alternately, they can be implemented digitally in software. One such system is described in U.S. Pat. No. 5,612,674 entitled High Sensitivity Apparatus and Method With Dynamic Adjustment for Noise assigned to the Assignee hereof and incorporated by reference herein.

While known approaches do provide a vehicle for suppressing or reducing uncorrelated noise in signals from ambient condition detectors, they also introduce delays. In the event that the parameter of interest, such as level of smoke or ambient temperature, does start to increase, the increases are attenuated and only appear in the output filtered signals after a delay interval which is characteristic of the type of averaging or filtering which is used.

As disclosed and claimed in the parent hereto, hereby incorporated by reference, a smoothing or filtering function can be altered in the presence of a fire indicating profile. It would be desirable to be able to start to adjust the degree of smoothing or filtering even before a fire indicating profile can be detected.

An apparatus in accordance with the present invention incorporates at least first and second ambient condition sensors. For example, and without limitation, a first sensor can be responsive to ambient smoke while a second sensor can be responsive to gas such as carbon monoxide or carbon dioxide or other hydrocarbons. The second sensor could alternately be responsive to temperature.

Control circuitry processes an output from the first sensor by filtering or smoothing same to reduce transients and or noise therein not correlated with the ambient condition to which the first sensor responds. The smoothing or filtering characteristics are then altered in response to an output from the second sensor. The control circuitry can be located adjacent to the sensors. Alternately, the control circuitry can, in part, be adjacent to the sensors and, in part, remote therefrom.

If both sensors are fire sensors, then one fire sensor could be used to control the response of the other fire sensor. Alternately, they effectively can interact by setting smoothing coefficients of each other.

If one sensor is a fire sensor and another is a non-fire sensor, then the non-fire sensor can be used to control the response of the fire sensor by slowing down its response. For example, the fire sensor may be responding to a false condition such as dust or humidity. Hence, if the first sensor responds to smoke (ion or photo-type sensor), the second could respond to gas, dust, temperature or humidity.

This is a combination detector where two or more sensors are logically combined in a detector to determine if a specific ambient condition exists. This ambient condition may be a fire, for example.

By controlling the smoothing coefficient, the detector is not prevented from alarming if a condition prevails long enough. It tends to be fail safe. The smoothing coefficient simply controls the rate of response and the development of a condition profile at any time will immediately increase the response from a non-profile condition because the detector stays active.

A method in accordance with the invention uses one or more of the sensors to establish at least one smoothing coefficient for another one or more of the sensors. The multiple sensors are used to improve the recognition of a condition or the absence of that condition in order to provide a faster response to the condition or a slower response to false conditions to prevent unwanted alarms.

Filter altering outputs from the respective sensors can be evaluated by amplitude, rate-of-change, or other means to determine if the condition is present. Processing can be carried out locally at the respective detector or remotely at a common control unit. In the latter instance, the coefficient(s) could be downloaded after having been modified if filtering is to be carried out at the detector.

The indication of a fire can be categorized or continuously based upon the sensor output values and evaluation of the values. For example, the indications can be categorized into:

No indication (there is no pattern or trend);

Slight indication (there is no pattern but a trend is started or there is a sharp increase to saturation);

Moderate indication (there is a pattern but significant doubt still exists about the condition); and

Strong indication (there is a specific pattern and little doubt exists about the condition).

Multiple sensors provide additional information beyond that available from a single sensor. As a result, a condition signature can be established to increase confidence that an actual fire condition is developing. This method is not limited to two sensors but can also include three or more.

Various types of filtering can be used and modified without departing from the spirit and scope of the present invention. For example, running averages can be used and dynamically modified in response to the output of the appropriate sensor.

The entire process can be carried out locally at each respective detector. Alternately, part of the processing can be carried out at the detector and part at a common control unit.

Numerous other advantages and features of the present invention will become readily apparent from the following detailed description of the invention and the embodiments thereof, from the claims and from the accompanying drawings.

FIG. 1 is a block diagram of a system in accordance with the present invention;

FIG. 2 is a block diagram of a multiple sensor detector usable in the system of FIG. 1;

FIG. 3 is a block diagram illustrating an apparatus and a process for altering a smoothing or fitting process in accordance herewith;

FIG. 4 is a flow diagram of processing in accordance herewith where both sensors are responsive to parameters indicative of fire;

FIGS. 5A-5D illustrates detector response to various conditions;

FIG. 6 is a flow diagram of processing in accordance herewith where a first sensor is responsive to a parameter indicative of fire and a second sensor is a non-fire sensor;

FIG. 7A and 7B illustrates detector response to various conditions; and,

FIG. 8 is a block diagram of an alternate apparatus and method in accordance herewith.

While this invention is susceptible of embodiment in many different forms, there are shown in the drawing and will be described herein in detail specific embodiments thereof with the understanding that the present disclosure is to be considered as an exemplification of the principles of the invention and is not intended to limit the invention to the specific embodiments illustrated.

FIG. 1 illustrates an alarm system 10 which incorporates a common control unit 14. The system 10 could, for example, be part of a more extensive building control system. It could also include a plurality of linked control units 14.

The control unit 14 can further include a programmable processor 14a, associated memory 14b and interface circuitry 14c. The interface circuitry coupled the processor 14a, via a communication link 16 to a plurality of ambient condition detectors 18.

The link can be unidirectional or bidirectional. The link can be implemented using for example, cable, optically conductive fiber or any other wired or wireless link such as RF.

The members of the plurality 18, 18a, 18b . . . 18n each include one or more ambient condition fire sensors S1 such as smoke sensors, flame sensors, gas sensors or heat sensors. At least some of the members of the plurality 18 include a second ambient condition sensor S2. The sensors S2 can correspond to fire sensors, or non-fire sensors such as dust, humidity or velocity sensors.

The members of the plurality 18 communicate with the processor 14a, via the communications link 16. Signals, indicative of the sensed ambient conditions for each respective detector can be transmitted to unit 14. The signals can be analog or digital form without limitation.

A plurality of function generating output modules 19 is also coupled to the link. Modules 19 can be commanded to activate solenoids or other types of output devices as desired.

Processor 14a, in response to processing modules and instructions stored in memory 14b processes the returned ambient condition signals or indicators from one or more of the detectors in the plurality 18. In the presence of a predetermined alarm condition, a plurality of audible and/or visual alarm output devices 22 can be energized, via communication link 24, to thereby provide both visual and audible indicators of an alarm condition in the region being monitored.

An operator's counsel 14d can be used to provide a visual indication to an operator of an alarm condition. Alternately, alarm or related conditions can be logged in a printer 14e.

The indicators from the detectors 18 can be filtered or smoothed to minimize the effects of uncorrelated noise. The degree of smoothing can be varied in response to the probability of a fire being present.

In the presence of random variations in the indicators being received from the members of the plurality 18, characteristic of uncorrelated noise, a high level of filtering or smoothing can be implemented. This in turn limits the response time of the system.

Filtering or smoothing can be implemented in the control unit 14. Alternately, it can be implemented in one or more distributed devices 18 including smoke detectors, fire detectors, gas detectors or the like without limitation.

If on the other hand the indicators from the detectors 18 correspond to a fire profile, then the processor 14a can respond by either reducing or bypassing the level of smoothing or filtering thereby providing faster system response time.

As described below, the filtering or smoothing process can be carried out in hard-wired circuitry. Alternately, the process can be carried out by the execution of pre-stored instructions in the form of a control program stored in the memory 14b or at the respective detector. The prestored instructions can be executed to implement single or multiple stage exponential-like filtering or smoothing functions as well as to carry out software based alarm processing.

If desired, the raw signals or indicators from the detectors can be pre-filtered or pre-processed in accordance with the system described in U.S. patent application Ser. No. 08/522,599 filed Sep. 1, 1995, entitled "Pre-Processor Apparatus and Method, " assigned to the assignee hereof and incorporated by reference herein. Other forms of preprocessing can be used without limitation.

By way of example, and without limitation, a representative dual sensor detector, such as detector 18a, is illustrated in FIG. 2. The detector 18a includes a housing 25 which carries a first sensor 26a and a second sensor 26b. The sensors 26a and 26b are coupled to the local detector control circuitry 26c.

It will be understood that the processing described subsequently can be carried out by control circuitry 26c, at the respective detector, or alternately at the control unit 14 without limitation. Control circuitry 26c could be implemented, at least in part, with a programmed processor and local control instructions.

Control circuitry 26c is in bidirectional communication with interface circuitry 26d. Interface circuitry 26b is in turn coupled to the communications link 16 so as to place the detector 18a in bidirectional communication with control unit 14.

It will be understood that ambient condition sensor S1, 26a, could be implemented as a sensor responsive to a potential fire condition. Such sensors include smoke sensors, flame sensors, thermal sensors, and gas sensors.

The second ambient condition sensor S2, 26b, could be implemented as a fire sensor or as a non-fire sensor. Potential non-fire sensors include humidity sensors, dust sensors and velocity sensors.

FIG. 3 illustrates an exemplary block diagram of the present process and apparatus. It will be understood that the block diagram of FIG. 3 could be implemented in either hard-wired form or via the control program stored in the memory 14b which is in turn executed by the processor 14a. It could also be implemented at one or more of the respective detectors in hardware and/or software or both.

An output from sensor S1 can optionally first be pre-processed in element or step 28-1 (indicated in phantom in FIG. 3). The output from preprocessor 28-1, on line 30-1, is smoothed or filtered in module or step 32-1.

An output from sensor S2 can optionally first be pre-processed in element or step 28-2 (indicated in phantom in FIG. 3). The pre-processed output, on line 30-2, is in turn filtered or smoothed in element or step 32-2.

The outputs from the filtering processes A1(t) and A2(t) on lines 341 and 34-2, respectively, are then combined to produce a processed output function, on a line 38. The signal on line 38 is then an input to an alarm processing element or step 46. Inputs to the alarm processor 46 can include a plurality of thresholds 46a along with a plurality of slopes or gradients 46b.

The characteristics of the filtered signal on the line 38 can in turn be compared to one or more of the thresholds 46a and/or one or more of the slopes 46b by processor element or step 46. Alternately, more complex pattern recognition methods can be used.

In the event that an alarm condition is detected, a message can be displayed on the console 14b. Alternately, the alarm condition can be logged on the printer 14e. Further, the audible and alarm indicators 22 can be energized thereby alerting individuals in the region being supervised of the presence of the alarm condition.

It will be understood that neither the characteristics of the filters 32-1, 32-2, 36 nor the implementation thereof are limitations of the present invention. It will also be understood that a fire profile could appear on a transient basis. In this instance, the level of smoothing would be reduced or bypassed thereby decreasing the response time of the system 10. However, if the detected profile fades away or disappears, as would be the case of a transient smoke level, for example, the level of filtering or smoothing could again be increased restoring system 10 to its normal quiescent operating condition.

Alternately, in the event that the signal on the line 30-1 tends to return to a clear air level, indicative of the absence of a fire condition, the level of filtering or smoothing, for example, in the filters 32-1 and 32-2, 36 can be reduced or bypassed so as to enable the processed output signal on the line 38 to return to a corresponding filtered clear air value sooner than would otherwise be the case. In this instance, once the filtered signals return to their respective clear air values, the level of filtering can again be increased.

By way of further illustration of the characteristics and operation of the present apparatus and method, two examples will be discussed subsequently. Example 1 is directed toward performance characteristics of a detector, such as detector 18a wherein sensors S1 and S2 are both responsive to fire conditions. Subsequently, in Example 2, processing and response characteristics of another detector, such as detector 18b, which combines fire sensor S1 with a second non-fire sensor S2 will be described.

A first fire sensor S1 is combined with a second fire sensor S2 where S2 controls the smoothing coefficient of S1. S1 will initially set its own smoothing coefficient based upon a fire signature profile from its own values as given below and is modified by S2:

S1 Only:

No indication sets the smoothing coefficient to 0.025 (very slow response). Slight indication sets the smoothing coefficient to 0.10 (slightly responsive). Moderate indication sets the smoothing coefficient to 0.15 (moderately responsive). Strong indication sets the smoothing coefficient to 0.25 (strongly responsive).

S2 Modification of S1 Smoothing Coefficient:

No indication only increases the smoothing coefficient by adding 0.025. Slight indication slightly increases the smoothing coefficient by adding 0.10. Moderate indication moderately increases the smoothing coefficient by adding 0.15. Strong indication strongly increases the smoothing coefficient by adding 0.25.

FIG. 4 illustrates the steps of processing for Example 1 wherein both sensors S1 and S2 are responsive to indicators of fire.

In a step 102, outputs from sensors S1 and S2 are obtained. In a step 104, the respective sensor outputs can be pre-processed, elements 28-1 and 28-2. In a plurality of steps 106, coefficient values are established based on the likelihood that the pre-processed signal S1', on line 30-1, is indicative of a fire condition. In a plurality of steps 108, a coefficient value established in steps 106 is modified in accordance with the likelihood of the pre-processed output from sensor S2, line 30-2, indicating a fire condition.

In a step 110, the pre-processed output S1' from sensor S1 is smoothed or filtered, element 32-1. In a step 112, the pre-processed output S2' of sensor S2 is smoothed or filtered, element 32-2.

In a step 114, a processed signal is formed, line 38, which is a result of combining filtered outputs on the line 34-1 and 34-2 associated with both sensors. The combining step is discussed subsequently.

In a step 116, the processed output, line 38, is then either compared to one or more thresholds 46a or a plurality of processed outputs received from line 38 are compared to predetermined slopes 46b, alarm processing element 46, to determine if an alarm condition is present. If so, an alarm is then generated.

FIG. 5A illustrates in graphical form outputs from sensors S1 and S2. Neither output is indicative of a fire. Processed output signal on line 38 fails to indicate the presence of an alarm condition.

FIG. 5B illustrates in graphic form the processed output signal on the line 38 where the output of sensor S1 is saturated in the absence of any fire indication from sensor S2. The sensor S1 output had not previously suggested a fire profile. It could have become saturated due to humidity, dust or a bug. If so, it may be a false indication.

As is illustrated in FIG. 5B, the line 38 signal increases linearly in response to the saturation condition of sensor S1 producing ultimately an alarm condition. Hence, the process tends to be fail safe.

FIG. 5C illustrates in graphic form the effect on the line 38 output signal where sensor S1 has saturated due to a slight indication of the presence of smoke and a strong indication of the presence of carbon monoxide received from sensor S2. As a result of the strong indication provided by sensor S2, the line 38 output signal reaches an alarm state faster than is the case in FIG. 5B. Hence, as a result of combining outputs from two different fire related sensors, an alarm condition can be indicated sooner than is the case with just a single sensor while still minimizing false alarms.

FIG. 5D illustrates in graphical form the positive benefits achieved by the present invention. There, a strong indication of fire from sensor S1 in combination with a strong indication from sensor S2 produces an output which goes into alarm relatively immediately as a result of the smoothing process having been adjusted by the output from sensor S2.

A first, fire, sensor S1 is combined with a second, non-fire, sensor S2. Sensor S2 controls the smoothing coefficient of S1. Sensor S1 will initially set its own smoothing coefficient based upon fire signature profile from its own values as given below and is modified by S2 to reduce the response if a non-fire condition is detected:

S1 Only:

No indication sets the smoothing coefficient to 0.1 (slow response). Slight indication sets the smoothing coefficient to 0.25 (slight response). Moderate indication sets the smoothing coefficient to 1.0 (strong response).

S2 (non-fire) Modification of S1 Smoothing Coefficient:

No indication does not modify the smoothing coefficient. Slight indication divides the smoothing coefficient by 2 (less responsive). Moderate indication divides the smoothing coefficient by 5 (much less responsive). Strong indication divides the smoothing coefficient by 10 (very much less responsive).

FIG. 6 illustrates steps of a method 200 wherein a representative detector, such as a detector 18b includes a fire sensor S1 and a non-fire sensor S2. Steps of method 200 which are the same as respective steps of method 100 have been assigned the same identification numerals. Those steps do not need further discussion.

In a plurality of steps 206, the value of a smoothing or filtering coefficient is adjusted in accordance with the degree to which the pre-processed output S1' from the sensor S1 indicates a fire condition. In a plurality of steps 208, the value of the smoothing or filtering coefficient established in the steps 206 is adjusted in accordance with the degree to which the pre-processed output S2' from the sensor S2 is indicative of the respective non-fire condition. Subsequently, using the value established for the smoothing or filtering coefficient, the respective sensor outputs are smoothed, steps 110 and 112, combined in step 114, and then processed to determine if an alarm condition is present, step 116. It will be understood that different smoothing coefficients could be used with different types of sensors.

FIG. 7A illustrates in graphical form the interaction between fire sensor S1 and humidity sensor S2 where a fire condition is being reported by sensor S1 but no fire condition is being reported by sensor S2. Conversely, the graphs of FIG. 7B illustrate the damping effect that sensor S2 has in the presence of a strong indication of the respective condition so as to minimize the likelihood of a false alarm being produced in a response to the signal from sensor Si.

With respect to the flow diagrams of FIGS. 4, 6, once the smoothing coefficient K has been determined, the sensor data sample values S1' and S2' can be smoothed. This process is dynamic and K may change with each new data sample. Recall that:

This method of determining K which is used for both sensors S1, S2 is not necessarily restricted. Different K values can be determined for each sensor [i.e. K1 for use in A1(t) and K2 for use in A2(t)).

For example, K2=(K1)/2 so A2(t) will have a slower response than A1(t). The relationship between K1 and K2 will be a function of the type of sensors involved and their response characteristics. To continue with the presently preferred method, K1=K2.

The processed individual sensor values A1(t) and A2(t) are now combined as in step 114 and used in an alarm determination routine, step 116 to arrive at a determination (Alarm or No Alarm). This function is expressed as:

F(t) may equal A1(t)+A2(t) as a straight sum, or A1(t)2 +A2(t)/2 as sum of squares, or A1(t)+A2(t)/2 as a weighed sum, or other mathematical relationship, without limitation, between A1(t) an A2(t).

Also, for the purpose of this invention, the actual equation for A1(t) and A2(t) may vary and be within the scope of determining the smoothing coefficient(s) by using more than one sensor. The presently preferred method is to let F(t)=F(A1(t),0]=A1(t) where A2(t) is just used to determine the smoothing coefficient K. This is the method demonstrated in the response graphs.

If A1(t)>alarm threshold AT, THEN OUTPUT=ALARM, step 116.

It will be understood that, alternately, the output signal from sensor S2 can be used to directly alter one or more coefficients of the smoothing process for the output signal of the sensor S1. In this regard, some or all of the processing of the S1's output signal can be carried out locally at the respective detector. the partially or fully processed output signals, could be analyzed at the detector and/or at the control unit 14 for the presence of a fire condition.

FIG. 8 illustrates an apparatus and process wherein an output from sensor S2 has been pre-processed in element or in step 28-2. That output, coupled by interface circuits 40 to circuitry 42, is intended to alter one or more coefficients of a two-stage filter of signal S1', illustrated as exemplary only. A first filter section 44-1 with a filtered output on a line 46-1 is in turn coupled to a second filter section 44-2 which in turn generates an output on a line 46-2. The twice-filtered output on the line 46-2 can then be analyzed in the alarm processor 46.

It will be understood that some or all of the elements illustrated in FIG. 8 could be located within the respective detector. Alternately, alarm processor 46 could, for example, be located at common control element 14. In yet another alternate, filters 44-1, 44-2 and altering circuitry 42 could also be located at control element 14. It will be understood that neither the way in which filtering is implemented nor the way in which it is modified in response to output from another sensor are limitations of the present invention.

It will also be understood in such an instance that control circuits, such as 26c, located at the respective detector could be implemented using a programmed processor. Preloaded instructions associated with that processor could be used to carry out the pre-processing, elements 28-1, 28-2 as well as filtering functions such as filters 44-1 and 44-2.

From the foregoing, it will be observed that numerous variations and modifications may be effected without departing from the spirit and scope of the invention. It is to be understood that no limitation with respect to the specific apparatus illustrated herein is intended or should be inferred. It is, of course, intended to cover by the appended claims all such modifications as fall within the scope of the claims.